WORKSHOP 02 · APRIL 18, 2026
Verifying STEM AI Outputs:
What You Really Need to Know
The Check the Machine protocol and error-hunting skills for AI-generated code and content.
April 18, 2026 · 8:30 AM – 12:00 PM PST · Virtual
CRAFT PD Series · UCF DRACO Lab & School of Teacher Education
The CRAFT Cycle
Today's session IS a CRAFT cycle. You'll experience it — then learn to build your own.
CContextualizeWhy it matters — real-world framing
RReframeSurface the wrong model → install the right one
AAssembleI Do → We Do → You Do
FFortifyVerify with tools + AI; errors are features
TTransferConnect forward to your classroom
8:45
C
Contextualize
Your students are already using AI — here's what we know
YOU DO
Contextualize
8:45 (15 min)
Self-Assessment: AI in Your Classroom
Answer honestly in the shared doc:
- How are your students using AI right now?
- What's your biggest worry about student AI use?
→ Open Shared Doc
9:00
R
Reframe
The problem isn't AI use — it's unverified AI use
LISTEN
Reframe
9:00 (15 min)
Poll + The Real Problem Isn't AI Use
Vote honestly — when a student uses AI: (a) plagiarism (b) didn't learn (c) might be wrong (d) can't tell
- Banning AI teaches avoidance.
- Teaching verification teaches engineering thinking.
- We don't ban calculators — we teach when and how.
YOU DO
Reframe
9:15 (15 min)
Breakout: Rewriting the AI Policy
Your group is an AI policy committee.
- Draft 3 classroom norms requiring VERIFICATION
- What does 'acceptable AI use' look like in your subject?
BREAK
9:30 (15 min)
Break #1
Spot the error in the AI 'fun fact' in chat.
9:45
A
Assemble
Error hunting — code first, then content
YOU DO
Assemble
9:45 (30 min)
I Do + Breakout: Code Error Hunt
First, watch us debug a seemingly-correct AI-generated Python function. Then in your breakout: 3 AI snippets. For each:
- Read it. Predict. Find the error. Fix it.
- Document findings in shared doc
- First group to catch all 3 wins!
→ Open Code Snippets
YOU DO
Assemble
10:15 (15 min)
Solo: Break Your Own AI
Ask an LLM to solve a problem from YOUR subject.
- Try to break it
- Document one error (or how you verified correctness)
YOU DO
Assemble
10:30 (30 min)
I Do + Breakout: Content Error Hunt
First, watch an LLM give a confident, subtly-wrong STEM explanation. Then in your group, use the Content Audit Checklist:
- Check claims, verify facts, assess grade-level
- Spot misleading simplifications
- Flag the sneakiest error for share-out
→ Open Audit Checklist
BREAK
11:00 (15 min)
Break #2
Preview the CtM one-pager.
11:15
F
Fortify
Check the Machine: Task → Before → After → Takeaway
YOU DO
Fortify
11:15 (30 min)
Follow Along: Live CtM Cycle
Run the same prompt. Write your 'Before' prediction. Compare 'After' to ours.
YOU DO
Fortify
11:15 (30 min)
Partner Practice: Run Your Own CtM
Pair up. Full cycle each:
- Task → Before → After → Takeaway
- Partner checks: 'Did you verify or just skim?'
→ CtM Template
YOU DO
Fortify
11:15 (30 min)
Build: Customize CtM for Your Classroom
Adapt the template for your students:
- Adjust language for grade level
- Pick a sample task from your subject
- Write student-facing instructions
11:45
T
Transfer
Building a verification culture
LISTEN
Transfer
11:45 (5 min)
CRAFT Debrief: You Just Lived It
Every activity today mapped to C→R→A→F→T — you ran two nested CRAFT loops: code and content.
- C — self-assessment + stakes
- R — poll + policy rewrite
- A — code & content error hunts
- F — Check the Machine protocol
- T — commitment + transfer to your classroom
YOU DO
Transfer
11:50 (10 min)
Pair-Share Commitment + Post-Survey
With a partner:
- One concrete step for verification-first culture
- 'I will...' not 'I might...'
- Share the boldest commitments with the group
- Then complete the post-survey before you sign off
→ Open Post-Survey
Your Resource Kit
Everything is yours. Bookmark, download, remix.
🤖 CtM Template
Editable Check the Machine protocol
🔍 Content Audit Checklist
Claims, sources, accuracy, grade-level
🐛 AI Error Gallery
Annotated AI failure examples
📄 Prompt Library
Prompts for verify-able outputs
📝 Post-Survey
Help us improve
Your students don't need to be
AI experts.
They need to be AI skeptics.
That's a skill you just learned to teach.
CRAFT PD Series · UCF DRACO Lab & School of Teacher Education
Dr. Mike Borowczak & Dr. Andrea Borowczak